distributed tensorflow demo; after https://databricks.com/tensorflow/distributed-computing-with-tensorflow

cluster_test.py

import tensorflow as tf

# run this script from the master node

cluster = tf.train.ClusterSpec({"distributed": ["192.168.102.90:1111", "192.168.102.90:1111"]})

x = tf.constant(2)


with tf.device("/job:local/replica:0/task:0/device:GPU:0"):  #"/job:distributed/task:1"):
    y2 = x - 66

with tf.device("/job:local/replica:0/task:1/device:GPU:0"):  #"/job:distributed/task:0"):
    y1 = x + 300
    y = y1 + y2


with tf.Session("grpc://localhost:1111") as sess:
    result = sess.run(y)
    print(result)

start_node.py

import sys
task_number = int(sys.argv[1])

import tensorflow as tf

# use this script to start the nodes
# for the first node: 
# python start_node 0


cluster = tf.train.ClusterSpec({"local": ["192.168.102.20:1111", "192.168.102.90:1111"]})
server = tf.train.Server(cluster, job_name="local", task_index=task_number)

print("Starting server #{}".format(task_number))

server.start()
server.join()